When Randomization Assigns a Starting Strategy: An Estimand-Led Framework for Frozen Embryo Transfer Trials

Randomized trials may assign a treatment strategy rather than a fixed, uninterrupted exposure. Frozen embryo transfer (FET) provides a clear example: women can be randomized to start natural, modified natural, or artificial endometrial preparation, while cancellation and rescue artificial preparation change the care subsequently received. Unless the estimand is explicit, a terminal livebirth comparison can be mistaken for a comparison of uninterrupted protocols. Using the randomized trial by Ho et al. as a worked example, this Analysis examines how intercurrent events - particularly cancellation, delayed transfer, and rescue switching - shape the clinical question and interpretation. It distinguishes a starting-strategy estimand that uses a treatment-policy strategy for these events from completed-protocol and hypothetical no-switch estimands; it also develops patient-centered efficacy variables and denominator-aligned safety estimands. The framework recommends that FET trials prespecify the primary estimand, report cancellation and switching as core outcomes, include time and burden, align safety outcomes with biologically meaningful denominators, and document operational rules. Here, strategy denotes the intervention assigned at randomization, whereas pathway denotes the care sequence realized after assignment. This FET framework may serve as a prototype for other randomized trials of treatment algorithms, but generalization requires setting-specific estimand and intercurrent-event specification.